A Refined Case Based Genetic Algorithm for Intelligent Route Optimization

摘要

This paper introduces a refined knowledge based approach to make Genetic Algorithms (GA) more effective. A concept called Case Based Reasoning (CBR) is combined with GA. This technology uses former solutions to the problem for deriving new solutions. The contribution of this CBR combination is a Case Base Maintenance technology that cares for both diversity and fitness in population. Former solutions namely, former problem's GA's solutions whose human/culture related problems are modified by human (expert) support, is maintained, selected, and used as GA's initial population for new but usually more than 80% similar problems. This technology weakens the factor "accidence" in favor of fitness and diversity balanced search especially towards human/culture related optimality. Instead of hoping for "accidentally good solutions" such technologies create "intentionally good solutions". Also, the Case Base (CB) combined with a new NI (Nearest Insertion) is used as a way to include/maintain human/cultural related knowledge into GA, which is called CBHOGA (Case Based Human Oriented GA). Human related knowledge is hard to formally express and mostly maintained by human support. This knowledge is kept in the Case Base and inherited as the initial population. Further, NI operations maintain this knowledge or preserve important parts of the solution in the initial population particularly by specifying insertion locations in the above mentioned new NI. The effects are shown by experiments.